xref: /freebsd/sys/dev/syscons/plasma/fp16.c (revision 95ee2897e98f5d444f26ed2334cc7c439f9c16c6)
1 /*-
2  * Copyright (c) 2015 Dag-Erling Smørgrav
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 #ifdef _KERNEL
28 #include <sys/libkern.h>
29 #else
30 #include <stdio.h>
31 #include <strings.h>
32 #endif
33 
34 #include "fp16.h"
35 
36 /*
37  * Compute the quare root of x, using Newton's method with 2^(log2(x)/2)
38  * as the initial estimate.
39  */
40 fp16_t
fp16_sqrt(fp16_t x)41 fp16_sqrt(fp16_t x)
42 {
43 	fp16_t y, delta;
44 	signed int log2x;
45 
46 	/* special case */
47 	if (x == 0)
48 		return (0);
49 
50 	/* shift toward 0 by half the logarithm */
51 	log2x = flsl(x) - 1;
52 	if (log2x >= 16) {
53 		y = x >> (log2x - 16) / 2;
54 	} else {
55 #if 0
56 		y = x << (16 - log2x) / 2;
57 #else
58 		/* XXX for now, return 0 for anything < 1 */
59 		return (0);
60 #endif
61 	}
62 	while (y > 0) {
63 		/* delta = y^2 / 2y */
64 		delta = fp16_div(fp16_sub(fp16_mul(y, y), x), y * 2);
65 		if (delta == 0)
66 			break;
67 		y = fp16_sub(y, delta);
68 	}
69 	return (y);
70 }
71 
72 static fp16_t fp16_sin_table[256] = {
73 	    0,	  402,	  804,	 1206,	 1608,	 2010,	 2412,	 2814,
74 	 3215,	 3617,	 4018,	 4420,	 4821,	 5222,	 5622,	 6023,
75 	 6423,	 6823,	 7223,	 7623,	 8022,	 8421,	 8819,	 9218,
76 	 9616,	10013,	10410,	10807,	11204,	11600,	11995,	12390,
77 	12785,	13179,	13573,	13966,	14359,	14751,	15142,	15533,
78 	15923,	16313,	16702,	17091,	17479,	17866,	18253,	18638,
79 	19024,	19408,	19792,	20175,	20557,	20938,	21319,	21699,
80 	22078,	22456,	22833,	23210,	23586,	23960,	24334,	24707,
81 	25079,	25450,	25820,	26189,	26557,	26925,	27291,	27656,
82 	28020,	28383,	28745,	29105,	29465,	29824,	30181,	30538,
83 	30893,	31247,	31600,	31952,	32302,	32651,	32999,	33346,
84 	33692,	34036,	34379,	34721,	35061,	35400,	35738,	36074,
85 	36409,	36743,	37075,	37406,	37736,	38064,	38390,	38716,
86 	39039,	39362,	39682,	40002,	40319,	40636,	40950,	41263,
87 	41575,	41885,	42194,	42501,	42806,	43110,	43412,	43712,
88 	44011,	44308,	44603,	44897,	45189,	45480,	45768,	46055,
89 	46340,	46624,	46906,	47186,	47464,	47740,	48015,	48288,
90 	48558,	48828,	49095,	49360,	49624,	49886,	50146,	50403,
91 	50660,	50914,	51166,	51416,	51665,	51911,	52155,	52398,
92 	52639,	52877,	53114,	53348,	53581,	53811,	54040,	54266,
93 	54491,	54713,	54933,	55152,	55368,	55582,	55794,	56004,
94 	56212,	56417,	56621,	56822,	57022,	57219,	57414,	57606,
95 	57797,	57986,	58172,	58356,	58538,	58718,	58895,	59070,
96 	59243,	59414,	59583,	59749,	59913,	60075,	60235,	60392,
97 	60547,	60700,	60850,	60998,	61144,	61288,	61429,	61568,
98 	61705,	61839,	61971,	62100,	62228,	62353,	62475,	62596,
99 	62714,	62829,	62942,	63053,	63162,	63268,	63371,	63473,
100 	63571,	63668,	63762,	63854,	63943,	64030,	64115,	64197,
101 	64276,	64353,	64428,	64501,	64571,	64638,	64703,	64766,
102 	64826,	64884,	64939,	64992,	65043,	65091,	65136,	65179,
103 	65220,	65258,	65294,	65327,	65358,	65386,	65412,	65436,
104 	65457,	65475,	65491,	65505,	65516,	65524,	65531,	65534,
105 };
106 
107 /*
108  * Compute the sine of theta.
109  */
110 fp16_t
fp16_sin(fp16_t theta)111 fp16_sin(fp16_t theta)
112 {
113 	unsigned int i;
114 
115 	i = 1024 * (theta % FP16_2PI) / FP16_2PI;
116 	switch (i / 256) {
117 	case 0:
118 		return (fp16_sin_table[i % 256]);
119 	case 1:
120 		return (fp16_sin_table[255 - i % 256]);
121 	case 2:
122 		return (-fp16_sin_table[i % 256]);
123 	case 3:
124 		return (-fp16_sin_table[255 - i % 256]);
125 	default:
126 		/* inconceivable! */
127 		return (0);
128 	}
129 }
130 
131 /*
132  * Compute the cosine of theta.
133  */
134 fp16_t
fp16_cos(fp16_t theta)135 fp16_cos(fp16_t theta)
136 {
137 	unsigned int i;
138 
139 	i = 1024 * (theta % FP16_2PI) / FP16_2PI;
140 	switch (i / 256) {
141 	case 0:
142 		return (fp16_sin_table[255 - i % 256]);
143 	case 1:
144 		return (-fp16_sin_table[i % 256]);
145 	case 2:
146 		return (-fp16_sin_table[255 - i % 256]);
147 	case 3:
148 		return (fp16_sin_table[i % 256]);
149 	default:
150 		/* inconceivable! */
151 		return (0);
152 	}
153 }
154